A traditional control system can basically be seen as a controller providing a control signal to one or more actuators affecting a physical entity, such as a machine or process, wherein one or more sensors determine actual conditions of the physical entity and provide feedback signals to the controller which compares, using a more or less complex control model, the conditions with desired conditions to provide the control signal with aim to make the actual conditions agree with the desired conditions. Sometimes, it is desirable to provide remote control via a network, e.g. a wireless network.
Traditionally, distributed automation systems have been connected using physical wires but with advent of technologies such as WiFi, Bluetooth, ZigBee, WirelessHART etc., wireless control have become commonplace. An important factor for automation systems is latency and with cellular systems of fourth generation coming down to round trip delays at around 20-40 milliseconds cellular technologies are becoming an interesting alternative that in addition to existing solutions offer better range, mobility and centralized management. FIG. 1 shows a distributed control system 100 where a sensor 102 and an actuator 104 are connected to a control node 106 over a network 108 for performing control of a process 110. The performance of the system 100 is highly dependent on information about communication delays 112, 114. If the network 108 is wireless the system 100 must also handle the case when connectivity is lost.
FIG. 2 shows a hierarchical control system 200 with a local controller 202 and a supervisory controller 204. This setup is common in for example robotics and the process industry. The local controller 202 may be a standard PID (Proportional Integral Derivative) controller and the supervisory controller 204 provides set point values. In the robotics case the trajectory may be remotely generated and while the control loops for the individual joints execute locally. Another example is coordination of autonomous aerial vehicles, as disclosed by A. Bemporad and C. Rocchi in the article “Decentralized Hybrid Model Predictive Control of a Formation of Unmanned Aerial Vehicles”, where the supervisory controller 204 runs a model predictive controller (MPC) for generating trajectories, mission planning and handle collision avoidance and the actual flight control system (stabilization, etc.) runs locally by the local controller 202.
Control systems may be arranged to operate in different modes and where each mode possibly corresponds to a different control strategy. A number of examples of hybrid control systems are disclosed in the PhD Thesis by Jörgen Malmborg, “Analysis and Design of Hybrid Control Systems”, Lund University. For example the hybrid controller in FIG. 3, which is represented as a Grafcet diagram with four states and two control modes. One mode is using an optimal approach (Opt Controller), e.g. when changing set point, and a PID controller (PID Controller) when the system output is close to the set-point. In the case of the autonomous aerial vehicles, as discussed in in the article “Decentralized Hybrid Model Predictive Control of a Formation of Unmanned Aerial Vehicles” by A. Bemporad and C. Rocchi, a hybrid control strategy may be applied to handle loss of connectivity.
From the discussion above, the need of an approach how to handle distributed control in view of connectivity is evident.